1. Forum
  2. Community
  3. General 3D Scanning Discussion
  4. How to improve real results with mechanical parts?

How to improve real results with mechanical parts?

  • Vaclav Zdrazil

    I'm getting pretty desperate. I had a little higher expectations for this scanner. 
    Do you have any ideas on how to improve the accuracy of the scan. I have already tried different angles and distances from the object. Even different methods for meshing, or different levels of noise reduction don't help much. 
    Does anyone have any advice on how to achieve the most accurate scan and the sharpest edges in the resulting scan. 


    I just want to scan the tip of this axis 



    The problem is that sometimes when I try to create a mesh, or remesh, the scanner freezes at this stage and then disconnects. It even disconnects if it is connected via LAN cable. It happens quite often. 



    Does anyone else have a problem with this too and how did you solve it. 



    When I give a refresh and the scanner reconnects, all my scans and projects disappear even though it shows that the memory is already partially full. The only thing that helps is to turn the scanner off and on. It's quite annoying.  





    Even when creating meshes, it often creates random artifacts that are not on the object or nearby. They don't even appear when the scan is displayed, only when the mesh is created, they appear out of nowhere. 

    I will be glad for any advice. I'm getting tired of this, because basically I haven't done a single successful scan yet to be satisfied with the result. 

  • TT

    Hi Vaclav,

    Thanks for this information, there are a few things happening here, so let me break them down and lay out some things you and Matter and Form can do to resolve these issues.

    1. How to get the most accurate scan and mesh. Accuracy is a function of the distance between the scanner and the object, so the closer you get, the higher the accuracy. THREE can do 33 microns at a distance of 220mm, and 400 microns at a distance of 700mm. That object is fairly small, so get close to it. Scan Density of Medium is likely to be good. You can try also try High Density but tbh I don't think it will vastly improve the scan for that particular piece. All that said, your base scan looks good to me, so I'll talk about your meshing issues next. 

    2. Meshing issues. The meshing is freezing at this stage? This is very odd. First of all, are you on an RC/nightly build? If so, please uncheck that and update the software to be the regular release version. This is the first we've heard about meshing freezing and the scanner disconnecting. There is a different (well, I think it's different, maybe it's the cause here too) error related to connection - an 'i2c' error - which we're currently fixing. These two may be related. Is it just the Flow Quad mesh with this problem, or is it happening for all of them? 

    3. After you refresh the scans are missing and memory is partially full, and the scan projects are missing. I expect this is related to the disconnetion problem. It's an error state and results in an error state. Once we sort out the disconnection problem, this problem will probably disappear I think. If not, we'll investigate it at well, of course.

    4. Artifacts during Poisson meshing - this is a normal part of Poisson meshing. It fills in areas that don't have data, and the bottom of the piece you've scanned, which you've trimmed, has a hole without point data, and it gets filled by Poisson. What you can do is after the mesh is complete, click the "Add to Project" button, then use the cleaning tools in the Scan screen to cut off the added part you don't want. 

    Another options for this piece is to not do the Merge / Remesh step, and just export the turntable scan from the Scans screen (the export is in the menu dropdown at the top left corner of the page). This is the raw point cloud meshed data, not reprocessed through Merge/remesh. It may be that this is just what you need and the merge/remesh step is unnecessary. It's worth trying and you can decide if the result is sufficent for your needs. The merge/remesh step cleans up the scan and is a more efficient mesh result, but if the raw point cloud data works for your need there's no reason to not use it.

    So, to recap: 1) please let me know what meshing disconnects, and if it's all fo the meshing options. 2) use poisson mesh, then add to project, and use the cleaning tools to delete the extra part it makes. 3) Just export the raw scan data from the Scans screen and see if that is sufficient for your needs.

    Thanks for flagging these issues! We'll get them sorted out and we're happy to help until you're satisfied with your scan results. 

    Trevor.@Vaclav Zdrazil 

    0 likes
  • Vaclav Zdrazil

    1. Scanned from various angles and distances ranging from 160-250mm from the subject. On medium and high density with different noise reduction (0, 15, 31) and the edges of the subject in the resulting scan still not to my liking.  

    2. I do not use "nightly releases". I have frontend version 6.8.6 and backend version 6.14.42.

    Yes the i2c error happens to me when scanning too.

    Freezing and disconnecting seems to happen randomly for all mesh modes. Even with just one turntable scan with 24 scans is often a problem. With higher scan counts it happens more often, so quietly Poisson Triangle on medium with 24 scans gets stuck in an endless loop and nothing happens for more than half an hour. Then I gave up. 

    Sometimes it gets stuck in the endless loop, even if I put cancel, and then I can't do much with the scanner.

    3. Yes it's related to disconnection, but I don't understand the reason for the disconnection, even when I have the scanner connected directly to the PC LAN cable it happens. I have also tried two different LAN cables. 

    4. Yes it also occurred to me to add them to the project and then edit them there, but I was rather surprised and didn't know it was common. 

    I'll try it again this weekend. I don't have much time during the week. @Trevor Townsend 



    0 likes
  • TT

    Thanks for this information Vaclav. Would you please contact support from within the THREE software? It will attach your log files, so we can take a look at them and determine what is the cause of your disconnection. 

    Regarding the edges of your scan, just to be sure, you are focusing the cameras on the object before you scan, correct?

    I would like to see up-close screenshots of the scan, both before you mesh, and after you mesh. From what I can see in the screenshots you posted, the scan looks pretty good, but I want to have a closer look to determine if it can be improved. 

    Trevor.

    0 likes
  • Vaclav Zdrazil

    @Trevor Townsend thank you for the update, the I2C error disappeared and hasn't reappeared since the update. Anyway, I'll send you the log files. The last log files before I updated the scanner to a newer version and the last log files from today with the updated software.

    Yes, of course, I have the focus of the cameras aimed precisely at the part of the component that interests me the most. I wanted to use it as a reference object. It's a relatively simple hollow shaft with splines, but the splines are tapered, so I wanted to know exactly how the scanner would handle it.

    I tried various settings for the intensity of the scanner projector's brightness (25, 50, 75, 100). I tried different levels (0, 16, 31) of noise reduction. The object is scanned from a distance of about 170mm at an angle of approximately 45°, so all the edges should be perfectly visible. Surprisingly, it manages to scan some of them, but it can't handle others.

    Normally, the lighting I have in the room is around 140 lux. I also tried using a light with dimming and a light diffuser, but even with lighting around 250 lux, I still can't achieve the ideal condition. Mainly, it seems to me that even merging 24 scans takes a really long time. Sometimes it doesn't even happen and seems to get stuck.



    before mesh 



    after mesh









    The medium mesh is completely unusable, regardless of which meshing method I choose.









    What frustrates me the most is that when I start meshing. I've even made a table for it and I know it can handle 24 scans in under 3 minutes, but the problem is that sometimes it gets stuck and nothing happens after that. When meshing for 24 scans takes more than 15 minutes. I usually find that I can't do much with the scanner anymore and I have to turn it off and on again.

    0 likes
  • TT

    Hi Vaclav,

    We've had a look at your log files and think we know what's happening. To help us troubleshoot and fix the disconnection problem, it would be very helpful if you could download and send us this scan project. Please send it to us if you don't mind. You will need to link it to us, rather than send it in an email, because from your log files it looks like there are over 600 scans in the project at full scan density, so it will be extremely large. If you need us to provide a drive link for you, just let me know.

    I'd also like you to try the following scan experiment: First, reboot your scanner, then create a new project. Do one medium density turntable scan of this piece, with only 12 steps. Please spray the piece before scanning it. After the scan is complete, use the cleaning tools to clear away any errant points surrounding the scan, then mesh it at High using Poisson. Once you've done that, please send us the log files, and that project, so we can take a look at it as well.

    Thanks!

     

    Trevor.

    0 likes
  • Vaclav Zdrazil

    Hi Trevor,

    I finally got around to finding the time to try everything out again. I sent you the scans, including the original ones. Yes, one of the test scans contained a lot of scans because I was trying out different scanning settings and conditions to compare the results and find the best option.
    Anyway, the scanning result with 12 steps, mid scan density, and hi mesh is not any better.

    Before meshing

    After meshing

    The point is that this shaft can obviously be scanned with fewer steps, but I'm using it as a reference object to know what to expect from the scanner because more complex objects will then be a much bigger challenge.



    By the way, a small detail in the cleaning tool after scanning - the "s" keyboard shortcut for lasso selection doesn't work for you, and undo Ctrl+Z takes the Z press according to the keyboard layout, where Z is bottom left, and doesn't take into account at all that I have a QWERTZ keyboard layout.
    For the future, following Revopoint's example, I wouldn't mind adding Clip, Select Connection, Invert Selection, and Polygon Selection options to the cleaning tool.
    And some kind of audio tone after completing an operation like scanning or meshing would also be useful, since it takes so much time.

  • RB

    @Vaclav Zdrazil I'm in the same boat of having many time when trying to remesh were the system hangs and the only way is to power cycle the unit. The one thing I found was my linux system with 16gb memory and i7 with a older video card Juniper PRO [Radeon HD 5750] had issues with a 16 scan it would only display with 12. When using a newer higher end system it would display them. That has not stopped the crashing of those scan when trying to remesh at a higher mesh. I wish there was a value indicating the status not just a bar that just sit there.

    0 likes
  • Vaclav Zdrazil

    @ROLAND BOILLE Yeah, I totally understand you. I've jokingly started calling the scanner a "waiting simulator". Luckily, I'm relatively patient, but sometimes it really annoys me, especially when it gets stuck almost at the end of something and then nothing happens for x minutes.

     

    I also don't have the newest PC, but even with hardware acceleration enabled in the browser, it often doesn't help. This is my PC setup to which I usually connect the scanner via LAN cable:
    Motherboard: ASRock X99E-ITX/ac Mini ITX LGA2011-3
    Processor: Intel Core i7-5960X 3 GHz 8-Core
    Memory: Crucial Ballistix Sport LT 32 GB (2 x 16 GB) DDR4-2400 CL16
    Video Card: EVGA SC GAMING ACX 3.0 GeForce GTX 1080 8 GBC: drive: Samsung PM961 1 TB M.2-2280 PCIe 3.0 X4 NVME SSD
    D: drive: Seagate IronWolf NAS 10 TB 3.5" 7200 RPM
    G: drive: Samsung MZ-75E4T0B 4 TB 2.5" SSD

    Operating System: Windows 10 Pro version 22H2 build 19045.5011.
    I'm using Chrome Version 130.0.6723.92 (Official Build) (64-bit) which is installed on the C: NVME SSD.

    We'll see what the guys can conjure up with updates.

    In any case, when you give it time, some quite usable things come out of it, although I now prefer to export most things and work directly on the PC. It's significantly more comfortable and faster.
    Here's one of the recent scans that will hopefully help us save a part for a machine at our workplace.


    before mesh
    after mesh 



    Almost final version 



    Deviation of the model from the scan



    I'm afraid the team is falling behind the curve, and Revopoint's MetroX is going to leave them in the dust =/ 

    0 likes
  • AB

    @Vaclav Zdrazil and @ROLAND BOILLE

    Thanks for your feedback and suggestions. We agree that the scanner takes a long time (and occasionally fails) when working with large data sets and we are working on a few things that should hopefully ease this pain point.

    Firstly, we're doing some optimizations in the merging process that should free up additional memory and speed up both combining scans and remeshing them.
    Secondly, we're working on some additional documentation to show when high-density point clouds are beneficial versus when you can get near-identical results with significantly less data, and by extension quicker processing times.

    We can also see that the lack of information with the progress indicator is adding to the frustration, as it's often unclear how long things will take and whether the process is still active, albeit slowly, versus when it has failed entirely. We'll see what we can do to improve the UI/UX in this area.

    Specific to Vaclav's scans: One thing you can try for that small rod item is to add planar geometry, i.e. put a box in the scan to aid the alignment. You're seeing the results of minor misalignment and that often happens with small items; the easiest way to fix this is to literally put it on a box. We've encountered similar challenges scanning a diamond ring and miniatures and putting them on a platform (which you will later edit out) assisted the auto-alignment of the turntable scans. You'd then fill in the missing areas on top/bottom with single-shot scans as usual.

    0 likes
Please login to reply this topic!

Search

Forum statistics

Page views (24h):
17
Page views (30d):
864
Topics:
22
Posts:
84
Members:
23

Newest members

© 2024, Matter and Form Inc. | Powered by Xenforum